Multilevel Autoregressive Modeling of Interindividual Differences in the Regularity of a Process

Michael J. Rovine and Theodore A. Walls

Rovine and Walls propose integrating a time series analysis and multilevel modeling. The model they present uses a first order autoregressive (AR(1)) model as the 'level 1' equation of a multilevel model. They compare the individual parameters estimated in this model to those obtained from a dedicated ARIMA modeling program (SAS PROC ARIMA). In this model the AR(1) parameters (individual intercepts and slopes) are used to model daily alcohol consumption reports of moderate to heavy drinkers. They expand this model to include covariates that explain the heterogeneity of drinking slope estimates. Throughout, consideration of relationship between typical longitudinal modeling goals and those of this analytic strategy are highlighted. The authors conclude with a consideration of potential broader applications of this model and possible next steps in the development of this approach.